Sinaps Studio v2.3.2

Sinaps Studio 2.3.2

Suite publication access

• Fixed application tokens and the Test API Send action remaining unavailable after signing in again to the same Sinaps Cloud account.
• Publication access now follows the durable host/account identity instead of treating a renewed login session as a different account. Existing tokens, policies and execution history remain available.
• The outbound execution relay also recovers after the same-account sign-in without requiring the Suite page to be opened.
• A different account or host cannot inherit publication access. Explicitly disabled access stays disabled, and execution configuration changes still require review.

API compatibility

This update does not rename or replace the public /api/v1 endpoints. Existing HTTP and MCP request formats and asynchronous GET polling remain supported. Older desktop clients are not required to upgrade to keep using the existing server endpoints.

Tracing

RAG query and result inspection is preserved: new embedding-query spans record the query and generated embedding, while retrieval logs include the authorized returned chunks, metadata, identifiers and scores. Credential redaction and size limits remain in effect.

Windows update

Use SinapsStudioSetup.exe to update an existing installation. The normal upgrade workflow preserves projects, models and user profiles. Verify the download using SHA256.txt; agent-release.json describes the matching installer version and checksum.

Validation note

During an automated sample Squad run, the reviewer's Cloud RAG request returned HTTP 503. Token creation and the Test API contract request succeeded through the Suite UI. The cause of that response was not determined; it is not included among the fixes claimed by this release.

Sinaps Studio v2.3.1

Published / Suite requests

• Edit HTTP methods, request URLs, bodies, and pre-request/post-request JavaScript scripts. Drafts survive tab changes and Suite rerenders and can be exported as a Postman collection.
• Scripts execute with time and memory limits. Application tokens stay protected, and requests remain scoped to the selected publication on the configured Cloud host.
• Automatic tracking uses GET polling without repeating the execution POST. The supported synchronous pm API is a subset of the Postman runtime.

First sign-in and LLM evaluation

• AWS Console Light is the default Studio and editor theme for new and existing accounts. Users can select another theme afterward.
• Signed-in users can evaluate OpenAI, Google Gemini and Mistral for 30 days per user, with OpenAI as the initial option when no personal provider is selected. The server tracks the period across logout and reinstallation.
• Evaluation credentials stay in the encrypted server vault and are never included in the installer or returned to the browser. Evaluation supports the configured models, up to 100 requests per user per day, and bounded response lengths. Personal provider settings remain independent.
• When no provider is available, chat displays a setup prompt with a button opening LLM provider settings.
• The unfinished GPU Security navigation entry is hidden. Agent governance, guardrails and access controls remain enabled.

Windows download

Download SinapsStudioSetup.exe below. The normal upgrade flow preserves existing projects and profiles. SHA256.txt contains the installer checksum; agent-release.json contains version and installer metadata. See RELEASE_NOTES.md for details.

Website and download portal: https://sinapsdata.com.br

Sinaps Studio v2.3.0

Sinaps Studio 2.3.0 expands the RAG workspace and makes published Squads easier to integrate, test, govern, and observe.

Suite: contracts, API testing, tokens, and usage

• Publish an entire Squad or an individual agent with explicit input/output contracts and configured policies, tools, RAG permissions, and limits.
• Work directly inside each published component with Contract, Test API, Tokens, Usage, and Integration tabs.
• Use the new Postman-style workbench: method/URL bar, Send controls, request tabs and editor, a resizable request area, and responses below the request. Inspect JSON, raw content, headers, HTTP status, time, size, and contract-validation results.
• Run real asynchronous executions or follow them automatically with GET polling. Idempotency keys prevent duplicate work when retrying submissions.
• Export Postman Collection v2.1 and OpenAPI 3.1 JSON, including contracts, examples, HTTP routes, and MCP requests. Exported authentication variables are empty; tester tokens are never included.
• Create expiring application tokens, revoke or delete them, and preserve execution history and traces.

Execution responses and observability

Both POST submissions and GET results now include:

• datetime and squad;
• token_in, token_out, and total_token, accumulated from actual provider usage across agents and nested runs;
• execution_latency_ms, including queue time, and response_latency_ms, measuring server processing for the current HTTP response;
• consultation: { status_url, method: "GET" }.

The creation timestamp stays stable across polls, and completed execution latency stops increasing. Trace Explorer connects the execution to its agents, SQL queries, RAG retrieval, tools, errors, and delivery.

API migration: use consultation.status_url in place of the former top-level status_url. The HTTP Location header continues to identify the result endpoint.

RAG and SinapsSQL

• Organization knowledge bases with StudioSQL schemas and tables, a default catalog, and an initial example table.
• Ingestion from multiple documents, project folders, relational databases, APIs, pipeline stages, datasets, JSON, and webhooks. Reuse dataset schedules and the data orchestrator for recurring workflows.
• Resource validation from the source workbench before ingestion.
• Documents, chunks, embeddings, semantic search, vector graphs, and schema/table/base relationship graphs.
• Database and table storage sizes, update timestamps, record/vector counts, and authorized table or record deletion.
• SinapsSQL with syntax highlighting, wrapped queries, and independent response scrolling.
• Expiring RAG keys, revocation/deletion, selected-table access scopes, request usage, and authenticated external API access on a dedicated port.

Workspace and governance

• Agent profiles, dossiers, governance investigations, limits, and approvals connected to execution records.
• Approval decisions record the authenticated user and time.
• Compact navigation, remembered return routes from Trace Explorer, separate base/table scrolling, and more stable RAG loading layouts.
• Compact themed notifications with a direct Open action, improved native window edges and DPI behavior.
• Portuguese, English, Spanish, and Simplified Chinese; light and dark themes.

Download and update

Download SinapsStudioSetup.exe below for Windows x64. Run it to install or update Sinaps Studio. The update workflow preserves projects, models, settings, and local user data. Close active work when requested by the installer.

The installer includes the native application and base runtime. Optional model runtimes and model weights are installed separately when requested. Development profiles, credentials, private projects, and corporate datasets are excluded.

Verify the download with SHA256.txt. agent-release.json supplies the matching version, size, and checksum for MCP bootstrap installation. RELEASE_NOTES.md contains the full English feature notes and response-field reference.

Sinaps Studio v2.0.0

Sinaps Studio 2.0.0 brings a more compact workspace, direct access to the project Agenda and Token dashboard, consistent AWS theme highlights, and clearer RAG storage information.

Navigation and workspace

• Centered top navigation with icons and labels for Context, Database, Data Orchestration, Assistant, Hardware monitor, and Community feed. Administrative options follow your account permissions.
• Reorganized left navigation, including Projects, Connectors, Datasets, Models, Suite, Canvas, Squad, Agenda, RAG, and Token. Collapsed sidebars use less space.
• Resources, Python libraries, Traces, Security, Users, and Version control are grouped on the right, according to availability and permissions.
• Notifications sit beside the theme selector and display the unread count.
• A shorter header, the project name above the navigation, smaller sidebar profile photos, and no duplicate account label in the top bar.
• Stable menu spacing when resizing the window or using Windows display scaling, including 125%.
• Refined Graphite surfaces, gray icons and logo, restrained corners, and slightly smaller sidebar text. The RAM indicator is positioned near the bottom of the right sidebar.

AWS Console Light

• Primary actions use AWS orange with dark, readable labels.
• Active tabs, selection indicators, icons, flow cards, and other highlights use the orange palette throughout RAG, Agenda, and Token.
• The + New user button follows the same theme palette.

RAG schemas and storage

• The storage summary now follows the selected schema instead of showing the size of the entire database.
• Each schema card in the graph displays the combined size of its authorized tables, including data and indexes.
• Selecting All schemas shows the combined total for the visible catalog. Changing the schema filter updates the summary, graph, and table list together.
• Missing measurements display Unavailable, rather than zero or a partial total. Byte, KB, MB, and GB conversions remain available.

Project Agenda and Token

• Agenda opens the project's activities and resources directly, without the Squad sidebar. List, Calendar, Resources, filters, date ranges, and time zones remain available.
• Token provides a dedicated view of AI usage, with period and project filters, provider breakdowns, input/output totals, and CSV export.
• Returning to the workspace preserves navigation context and dashboard filters.

Account, terminal, and hardware monitor

• Fixed the Metadata limit is 64 KiB error when changing a profile photo. The Cloud API accepts validated JPG, PNG, or WebP images up to 2 MB.
• CPU charts display the latest sample without delayed interpolation and reuse their drawing surface between updates.
• Terminal startup is independent of optional assistant initialization, with failure handling and a retry action.
• Improved terminal resizing, tab selection, and file pasting so upload messages and late responses do not disrupt the terminal layout.

Installation and optional extensions

Download SinapsStudioSetup.exe from Assets and run it to update your Windows x64 installation. Finish active pipelines and Python sessions before updating. Project data remains outside the application directory.

PDF/chart reporting components and Conda can be downloaded and activated from Settings → Extensions. They are not downloaded automatically during installation or when opening that page. Model execution environments are prepared according to the selected Python environment and requested capabilities.

Verify the installer with Get-FileHash .\SinapsStudioSetup.exe -Algorithm SHA256 and compare the result with SHA256.txt attached to this release.

Learn more at Sinaps Data.

Sinaps Studio v1.5.0

Visit Sinaps Data.

October 31, 2026

This release improves readability, refines workspace layouts, and keeps database work connected across Studio and chat. New light themes, compact navigation, and refined chat controls make everyday work more comfortable. Saved connections now carry through resource discovery, DBMS exploration, and paginated queries in chat.

Installation

Download the installer from Assets below.

Platform	Filename	SHA256
Windows x64 · per user	SinapsStudioSetup.exe	816e7a9ff49e684889c074d1e725507685c64adcc6387b9922bfc9514f1bf309

Light themes and readability

• Four new light themes near the beginning of the theme picker: Lakehouse Light, Redis Red Light, Azure Blue Light, and AWS Console Light.
• Neutral light surfaces replace the warm cast of earlier iterations, with distinct accent colors for each theme.
• Consistent top bars for AWS, Azure, Redis, and Lakehouse, with clearer contrast and visual hierarchy.
• Larger interface text, lighter font weights, and improved contrast across navigation, panels, and controls.
• In light themes, Ingestor code cells use a light background that follows the graph canvas, with syntax colors suited to light surfaces.

Navigation that leaves more room for your work

• Expand or collapse the left and right navigation columns independently. The left column starts expanded and the right column starts collapsed.
• Expanded navigation shows resource names and fits its width to the labels, reducing unused space and reliance on hover tooltips.
• Show/hide controls use icons that match each column's position. Right-side icons stay aligned with the right edge.
• A wider, right-aligned RAM meter makes memory usage easier to read.

Data orchestration and cells

• The orchestration header now shows the page title and Back action in the Studio language, with Portuguese, English, and Simplified Chinese support.
• Removed the redundant Cell 1 → Cell 2 → Cell 3 strip to leave more room for the graph and steps.
• Fixed unintended borders inside cells and the focus outline that appeared on the first execution.
• Improved Python completion in Ingestor cells, both while typing and with Ctrl+Space. Local suggestions complement kernel completions and remain available while the kernel is busy.

Chat, agents, and Agent Laboratory

• A more compact chat header places related actions closer together and moves settings beside the message input.
• The drag handle and panel number appear when more than one chat is open.
• The model refresh button has a smaller footprint and a more restrained hover state.
• Light-theme message inputs no longer show the doubled blue outline or an unintended divider between text and controls.
• Closing and reopening chat within the workspace preserves its selected provider and model, including chats with no sent messages.
• Agent and Laboratory messages have more space from the edges. Their scrollbar sits closer to the right edge, with a thinner thumb and transparent track.

Saved connections and database queries in chat

• Opening a connector in Studio shows its saved connections. Search or refresh the list, create a connection, or return to the form without losing the draft.
• A header button returns to the saved list. Compatible database connections offer Open in Studio DBMS, selecting the correct connection while preserving query drafts.
• Remove is available for connections the user is authorized to manage. Confirmation identifies the connection and explains that its access grants will also be removed; existing queries and datasets are preserved.
• Chat can discover connections across connector types, assigned token metadata, installed datasets and models, and configured AI providers, according to the current account's permissions. Credentials and token values remain protected.
• When several connections match, the assistant can list their real names and ask which one to use. The selected connection carries into follow-up requests, keeping existing-database queries out of the connection setup form.
• The assistant can inspect schemas, tables, and columns on the chosen connection and ask for clarification when needed. Query results remain available as paginated tables in chat; DBMS can be opened on request for further exploration.
• Combined requests, such as opening DBMS and querying data, retain the task context through table or schema clarifications. Execution follows Studio's authorization flow.

Physical paths and file search

• File and folder tooltips, file tabs, and Copy path show the physical project path instead of the legacy unidade/... address.
• Search results emphasize the filename and extension, with the physical directory on a secondary line and the full path available on hover.
• Files, Commands, and Resources have separate groups, larger text, internal scrolling, and clearer distinctions between files with the same name.
• Filename matches take priority, recent files retain their visit order, and results support ↑/↓, Home, End, Enter, and Esc.
• Refined focus styling, sizing, and positioning keep search usable in light themes, dark themes, and smaller windows.

Updating an existing installation

Run SinapsStudioSetup.exe and choose the existing installation to update. The installer preserves projects, models, settings, and the local database. Allow active pipelines and Python sessions to finish before continuing if the installer requests it.

Validation

The distributable passed the native release audit, compiled-backend regression checks, and the installer contract tests. The updated database worker passed its import and stable-ABI checks. Interface changes were checked in light and dark themes, including compact windows and keyboard navigation.

Learn more at Sinaps Data.

v1.0.0

Sinaps Studio v8.0.38

Visit Sinaps Data.

Sinaps Studio brings AI models, datasets, assistants, agent teams, and community collaboration into one Windows desktop workspace. Explore the model catalog, coordinate work with Agent Squad, observe AI behavior in the Agent Laboratory, compare LLM responses, and share results through Community AI.

Version 8.0.36 corrects profile photos in People, Community, and call portraits, and brings together the latest Cloud account, organization licensing, invitation, pipeline, and call session fixes.

Installer hashes

Download the installer from Assets below.

Description	Filename	SHA256
Windows x64 · per user	SinapsStudioSetup.exe	55d54862ed32db36c2071ebfa355c2d0acb570543c66c39e2a6d1181e3996742

Highlights

• Community AI: Share models, datasets, training progress, metrics, and diagnostics in a collaborative feed.
• Model Catalog: Discover model families and variants, filter the catalog, review compatibility information, and download models into your workspace.
• Agent Squad: Plan work with agent teams, assign models and effort, and visualize roles, routing, and connections.
• Agent Laboratory 1:1: Let two configured agents discuss a topic, observe their responses, and join the conversation to guide the exchange.
• LLM Model Comparison: Send the same prompt to multiple configured models and compare responses, complete-response times, and reported token usage side by side.
• Your account, consistently displayed: My account, the profile panel, the Studio header, People, Community, and call portraits use the connected Cloud identity.
• Organization licensing: Members inherit their organization's license, while administrators manage subscriptions and payments.
• More relevant AI assistance: Ingestor cells, .sinaps cells, and the LLM terminal use instructions suited to their workflows.
• Clearer administration: Improved user registration and invitations, restored access to user management, and green feedback for successful invitations.
• Call session recovery: Improved handling of session refreshes during calls.

Community AI

Bring project results into the conversation. Community AI connects the organization feed, communities, shared files, and discussions with the work happening inside Studio.

• Publish result cards for models, datasets, training runs, metrics, and diagnostics, with attached files and snapshots.
• Follow shared training updates through progress, steps, and metrics such as loss, learning rate, and epochs.
• Explore shared results, join community conversations, and use comments, reactions, and sharing to keep the discussion connected to its artifacts.

Model Catalog

Discover Hugging Face model families and choose the variant that fits your workflow.

• Search model families or repository IDs and browse featured collections.
• Filter by task, parameter count, library, inference provider, and execution or hardware criteria.
• Review model details and available compatibility information, explore quantized options, and download the selected model or variant.

Agent Squad

Turn a task into a coordinated plan with specialized agents, model assignments, and a visual execution graph.

• Define workstreams and agent responsibilities, then assign providers, models, effort levels, and planned token budgets.
• Review and approve plan revisions before execution, with coordination and final review handled through the work plan.
• Configure routing and connections, and manage tools, memory, RAG context, and policies from the squad workspace.
• Inspect the agent graph and connected squads, and follow work through the board and operations views.

Agent Laboratory 1:1

Choose a topic and let two AI agents discuss it in alternating turns. Follow the conversation to observe how their roles and models shape their responses, how they develop arguments, and where they agree or challenge each other.

• Configure each agent's name, specialty, role, provider, and model.
• Set the topic and interaction limit, then start, pause, continue, or stop the discussion and adjust its pace.
• Join the conversation with questions or instructions to guide the next exchange.
• Save the conversation and export it as Markdown or JSON for later review.

LLM Model Comparison

Compare how configured LLMs respond to the same request without switching between separate conversations.

• Send a shared prompt to participating model panels and read their answers side by side.
• Compare time to the complete response, provider-reported output tokens, and relative response speed for completed requests.
• Share a comparison snapshot to the Community AI feed for discussion and follow-up.

Account and profile

• Fixed an incorrect administrator photo appearing beside your own name in People and Community. Photos now use the displayed person's account, with the default avatar shown when no photo is set.

• Corrected cases where a Cloud user saw the local technical account's name, administrator role, or profile photo.

• Profile details, password changes, photos, and session history now belong to the connected Cloud account.

• Account changes discard outdated responses and clear drafts, passwords, and photos from the previous account.

• Accounts without a profile photo display their own initials. Cloud errors no longer reveal a cached or local account profile.

Organization licensing

• Users and coordinators see their organization's license status without personal activation, subscription management, or trial purchase actions.
• Educational organizations display their educational license. Commercial organizations display the applicable license state.
• Professional trials retain their original 30-day deadline and show days used and remaining. Signing in or reinstalling does not restart the trial.
• Member access follows the organization's entitlement, including suspension, revocation, and trial expiration.
• Individual payment requests exclude users and coordinators; billing remains with the administrator.
• A logout option is available in the license dialog when access is blocked, allowing the user to switch accounts.

AI assistants and pipelines

• Ingestor assistance focuses on Python pipeline methods instead of defaulting to generic chart examples.
• Improved Python response handling in .sinaps cells and support for clarification requests.
• The LLM terminal better recognizes explicit read-only command requests and avoids unrelated navigation. Command execution still requires approval.

Collaboration and administration

• Improved call recovery when authentication sessions refresh, reducing incorrect session-change errors.
• Restored user management navigation after workspace changes.
• Fixed manual user registration for the invitation workflow without an initial password.
• Corrected invitation download links used by the Cloud email service.
• Successful invitation messages appear in green, failures remain red, and retries clear the previous feedback state.
• The administrative account directory now shows inherited organization licenses and identifies the administrator as responsible for billing.

Installation and upgrade

1. Download SinapsStudioSetup.exe from the release assets.
2. Run the installer and update your existing installation.